Research project The Foundations of Epistemic Normativity

Purpose and Aims
Epistemologists investigate norms of rationality and reasoning (how one ought to reason), as well as standards of justification (what one is permitted to believe given one’s evidence). They investigate fundamental epistemic values, such as knowledge and truth, and seek to articulate epistemic virtues, such as responsiveness to evidence. In everyday life, we make epistemological claims, for instance that one ought to believe in anthropogenic climate change, or that this belief is justified given the evidence. All this suggests that epistemology is normative—it has to do with norms, standards, obligations, values, reasons, virtues and permissions.

The idea that epistemology is normative raises fresh questions about the semantics, metaphysics and epistemology of epistemology. Do epistemological statements (e.g. ‘current evidence justifies belief in anthropogenic climate change’) represent how things are? Are there objective, epistemological facts? If there are, can we come to know them? These are key questions in metaepistemology, many of which carry over from meta-ethics, which has traditionally been concerned with normative ethical claims, such as that one ought not to kill. Meta-epistemology is a burgeoning new field that investigates such questions concerning the foundations of epistemological normativity. This project is at its vanguard.

Our purpose will be to defend robust realism about epistemological normativity: with regard to language and mind, we hold that (a) epistemological statements are representational, and (b) epistemological judgments are beliefs; with regard to the metaphysics of truth and reality, we hold that (c) some epistemological statements are true substantively and absolutely and (d) there are objective epistemological properties and facts; and with regard to epistemology, we hold that (e) it is possible to know these facts, and (f) this knowledge is substantive, and not merely procedural or instrumental.

Survey of the Field
Realism and anti-realism are old rivals in philosophy, and both morality and epistemology are among the most frequented battlegrounds. Anti-realists might reject some or all of (a)-(f). For instance, one well-known form of anti realism is relativism, according to which moral or epistemological truths are not absolute, but only relative to some perspective or standard. Relativism is highly influential in the humanities as a whole, particularly in science and technology studies, where relativist and social constructivist accounts of scientific knowledge dominate (Cf. Barnes & Bloor 1982; Latour & Woolgar 1986; Kusch 2002). Another prominent form of anti-realism is non-cognitivist expressivism, the view that normative statements and judgments do not represent, but express attitudes that are more like desires or preferences (Ayer 1936; Blackburn 1998; Gibbard 2003). Yet another is error theory, according to which normative statements are false (Mackie 1971; Olson 2014).

Though robust realism too has a substantial following (Boghossian 2006; Cappelen & Hawthorne 2009; Williamson 2015), the specific issues that arise from the idea that epistemology is normative have neither been fully explored nor systematically addressed. In taking seriously this view, our approach promises to shed new light on enduring problems, and opens up new avenues of investigation. In the following, we provide background to the three main areas of our inquiry: language and mind; truth and reality; and epistemology.

Survey: Language and Mind
Recently, there have been major advances in the defence of anti-realist theories of the meaning and truth of normative and evaluative statements. For instance, ‘new relativists’, such as MacFarlane (2014), Kölbel (2002), and Lasersohn (2016), have articulated extremely clear and sophisticated formulations of relativism about truth, presented as innocuous extensions of the influential possible worlds semantics developed by Lewis (1970) and Kaplan (1989). Expressivism has similarly undergone an overhaul, building on possible worlds semantics to provide sophisticated truth conditional semantic theories for normative sentences and judgments (Gibbard 2003; Yalcin 2007; Dreier 2009; Chrisman 2012; 2016). Others have developed hybrid expressivist theories according to which normative statements express both non-cognitive attitudes and beliefs (Boisvert 2008; Ridge 2006).

Some philosophers have applied relativist or expressivist semantic theories to epistemology (Gibbard 2003; Field 2009a; 2009b). Though there has been critical discussion of these semantic theories, critics have tended to focus on general issues, or issues that arise in other domains, rather than those that arise specifically with respect to epistemology (Schroeder 2008; Cappelen and Hawthorne 2009).

Our own view is absolutist: epistemological statements are factual, and some are true absolutely. The more detailed positive account of the semantics of epistemological language and thought that we will develop draws on another innovation in possible worlds semantics: the two dimensional semantic framework which develops the Fregean insight that there are two kinds of meaning—sense and reference—as well as two kinds of mental content—narrow and wide (Chalmers 2006; Jackson 1998). On our view, normative, epistemological concepts such as ‘justification’ have both a normative dimension and a factual dimension, whereby they pick out a real, normative property.

Survey: Truth and Reality
One powerful line of objection to robust realism stems from the adoption of the naturalistic view that all facts can be explained by natural science. Many philosophers who are tempted by anti-realism about normativity are persuaded in part by the perceived failure of normative naturalism. As G.E. Moore (1903/2003) famously claimed, normative concepts cannot be reductively analyzed, and as Mackie (1971) argued, if there were objectively normative facts, they would have to have ‘do-be-doneness’ built right into them, and hence would be ‘queer’. Considerations such as these, together with the idea that epistemology is normative, have led some anti-realists to reject (c) and (d) (Cf. Field 2009a; Gibbard 2003; Olson 2014).

Normative naturalists have responded to these arguments by postulating more sophisticated theories, according to which normative concepts need not be reductively analysed, but still pick out natural properties (e.g. Boyd 1988). However, normative naturalists have faced a litany of objections; one that is directly relevant here is that naturalism entails a kind of relativity (e.g. Horgan & Timmons 1991) Our view is more closely allied to Moore’s non-naturalism—the view that normative properties are sui generis, and cannot be explained in naturalistic terms. This approach to normativity has been gaining momentum in meta-ethics (Cf. Cuneo 2007; Shafer Landau 2003, Wedgwood 2007). Though some (Cuneo 2007; Wedgwood 2007) have considered the application of non-naturalism to epistemology, the view deserves further investigation.

Survey: Epistemology                                                                                                                                      The most developed discussion of knowledge of epistemological norms has tended to focus on logical norms and principles, such as Modus Ponens (hereafter MP): Q logically follows from P, and if P, then Q (Field 2009a, 2009b; Boghossian 2001; Phillie 2006; Enoch and Schechter 2006, 2008). Influenced by Ryle (1946) and Carroll (1895), the consensus is that robust realism about such norms or principles generates regresses or vicious circularity, so no such norm could ever be justified or known. Many also claim that acceptance of MP is not belief, but some non-cognitive attitude of some kind.

A related issue concerns whether our knowledge of epistemological norms is procedural knowledge – or knowledge how to reason – versus substantive knowledge of the norms of valid reasoning (Ryle 1946), and many think that knowledge of logical norms and principles is merely procedural (Field 2009a).

Our view is cognitivist: we hold that epistemological judgments are beliefs, and that epistemological beliefs can amount to knowledge. We will build on the idea that procedural knowledge may nonetheless be substantive (Stanley & Williamson 2001; Stanley 2011). This view has been applied to regresses similar to those about justification.

Theory
Language and mind
As robust realists, we hold that (a) epistemological statements serve to represent, (b) epistemological judgments are straightforwardly factual beliefs, and (c) some such beliefs are substantively and absolutely true. For instance, if you sincerely assert ‘current evidence justifies belief in anthropogenic climate change’, you represent a real relation between current evidence and belief. If belief in anthropogenic climate change is in fact justified by current evidence, then what you say is true, substantively and absolutely. If epistemology is normative, claims (a), (b) and (c) face significant opposition from non-cognitivist expressivists, relativists and error theorists. As we have seen, all three of these views have received a more sophisticated treatment and powerful defence recently (§2.1).

First, we will engage with these sophisticated anti-realist semantic theories, as they are applied to the epistemological domain. Do any of them give an adequate account of epistemological language and thought? We will argue that relativism and expressivism cannot make adequate sense of epistemological disagreement—about, for instance, whether belief in anthropogenic climate change is justified. A consequence of both relativism and expressivism is that the parties to the disagreement do not genuinely disagree—they merely talk past one another. We will expand these considerations to normative disagreement more generally, and take into account a wider range of anti-realist theories.

Second, we will defend the cognitivist view that judgments are beliefs. Noncognitivists argue that normative judgments are unlike ordinary factual beliefs because the former but not the latter are intrinsically motivating. In epistemology, some suggest that if you judge that MP is a valid form of reasoning, then you must be motivated to reas on with MP: if you believe P, and if P then Q, then you must be motivated to believe Q. (Blackburn 1995; Boghossian 2001; Carroll 1895; Ryle 1946) We will argue that epistemological judgments are not intrinisically motivating, with reference to related debates about the normativity of reasoning and the normativity of belief. As Harman (1986) has shown, you might accept MP, yet fail to be motivated to believe Q, even though you believe P and if P then Q. Moreover, as Glüer and Wikforss (2013) have argued, some norms, such as ‘you ought to believe P iff P is true’ do not normatively guide belief, and are not intrinsically motivating. We will argue, on the basis of these and similar cases, that epistemological judgments are not intrinsically motivating.

Third, we will give a positive account of the semantics of normative statements and judgments that both solves the problem of lost disagreement, and offers an alternative to the motivational internalist characterization of normative judgment. Building on the two-dimensional semantic framework, we will argue that the meaning of normative predicates can only be characterized in normative terms, providing an alternative to the account of normativity as motivation or guidance. Furthermore, we will argue that normative predicates pick out sui generis normative properties, and hence that there is a shared meaning in cases of disagreement, so that the problem of lost disagreement can be solved.

Truth and Reality
Robust realists hold (c) that there are some substantive and absolute epistemological truths and (d) that there are objective, epistemological properties and facts. For instance, it is true that current scientific evidence justifies belief in anthropogenic climate change, and this is true in virtue of objective, epistemological facts.

Our approach to (c) and (d) has two main elements. First, we will ask: in what sense is epistemology normative? Anti-realists often approach this question from the perspective of semantics and psychology, arguing, for instance, that normative judgments are intrinsically motivating. As we have already indicated (§3.1), this view does not obviously carry over to epistemology. Building on our two-dimensional semantic theory for normative concepts, we hold that there are objectively normative facts concerning such matters as whether one’s evidence provides reason for belief. On our view, epistemology is normative in the sense that epistemological truths represent these facts.

However, robust realism faces an important challenge. It appears that certain very plausible epistemological principles conflict with one another or face counterexamples. This is because in debates about epistemic, logical and rational normativity the focus has generally been restricted to universally quantified deontic principles, such as (T) ‘For all agents, S: S ought to believe that P iff P is true’ or (E) ‘For all agents, S:S is permitted to believe that P iff S has sufficient evidence that P’. (Bykvist & Hattiangadi 2013; Glüer & Wikforss 2013) Since (T) and (E) conflict whenever the evidence is insufficient or misleading, it seems we must choose between them, though both are plausible.

In response to this difficulty, we will explore the hypothesis that the normative, epistemological facts only concern the presence or absence of defeasible, normative, epistemological reasons. For instance, suppose you have evidence against a true proposition P. In this case, the truth of P gives you a normative reason to believe P, while your evidence against P gives you a normative reason to refrain. The question whether you ought to believe P, all things considered, is a further, normative question that is not exclusively epistemological. We will argue further that deontic principles such as (T) or (E) should be reformulated as generics, rather than universally quantified principles. A generic statement such as ‘chickens lay eggs’ is a true generalisation, even though male chickens do not lay eggs. If epistemic principles are understood to be generics, then they too are true in virtue of facts concerning defeasible reasons, even though they are not universal truths.

Second, we will investigate whether familiar objections to naturalism in metaethics apply in the epistemological domain. For instance, it has been suggested that the concept ‘justification’ can be naturalistically analysed, in terms of
evidence and probability, and thus that a Moorean open question argument against naturalism is not applicable here (Heathwood 2009; Lenman 2008). However, we will argue that other prominent objections to naturalism in metaethics do carry over to epistemology. One such objection shows that naturalism entails a kind of relativism, and a corresponding problem of lost disagreement (Cf. Horgan & Timmons 1991).

Third, we will defend non-naturalism in epistemology—the view that epistemological properties are sui generis and irreducible—against standard objections. For instance, a central version of Mackie’s queerness argument against non-naturalism depends on the suggestion that normative judgments are intrinsically motivating, an assumption we have independent reasons to reject in the case of epistemology.

Epistemology                                                                                                                                                   As robust realists we hold that (e) it is possible to know epistemological facts, and that (f) this knowledge is substantive, and not merely procedural. Again, we endorse a form of cognitivism about epistemological judgment, according to which epistemological judgments are beliefs, some of which amount to knowledge.

Anti-realists frequently object to (e), the claim that epistemological facts are knowable, on the basis of the wide variation in epistemological judgments across history, culture or ideology – for instance in judgments about what kind or degree of evidence or justification is required for a belief to be adequately supported, or what kind of logic is required for a piece of reasoning to be valid. The problem, according to anti-realists, is that there is no independent basis for adjudicating between these different judgments, and so no way of knowing whether a belief is justified or an argument is valid (Barnes & Bloor 1982; Kusch 2002, Rorty 1991). If so, they argue, there is no basis for the kind of cognitivism that the robust realist aims to defend.

A related anti-realist challenge is that we cannot objectively justify our reasoning methods because attempting to justify them would either be circular or lead to regress. Consider MP: one might think any justification for it would either be circular (you need MP to justify MP) or lead to a regress (any logical principle you took to justify MP would in turn have to be justified). If MP cannot be justified, then robust realism seems on shaky ground: it would be unwarranted to claim that we know MP to be valid. It thus seems that the explanation of why it might be in some sense reasonable to accept MP will have to proceed in an antirealist way. And indeed some anti-realists in logic have recently argued that we should think of MP as a nonfactual, perhaps instrumental norm or policy that agents endorse or accept, not for their correctness or validity but for their consistency with these agents’ plans or other commitments. (Field 2015; MacFarlane 2014).

We will argue that robust realism has the resources to vindicate the idea that we can have knowledge of justification or validity. Concerning variation, we will show that there is no straight argumentative route from the idea that there is no independent – ‘from nowhere’ – basis for adjudication between two views to the rejection of objective justification. We will appeal to types of justification that are not ‘from nowhere’, such as abductive justification, or inference to the best explanation, which requires no neutral vantage point.

Concerning regress and circularity in the justification of our principles of reasoning, such as MP, we will question first whether all kinds of objective justification lead to vicious regress or circularity: we will argue both that the threats of regress for cognitivism in the philosophy of logic are vastly overstated and that circularity is not necessarily vicious. In particular, we will look at how different types of justification might help vindicate robust realism, while being in some sense circular. We will in particular focus on the reflective equilibrium or abductive method of justification of logical principles, which enjoys a kind of renewal in the epistemology of logic (Goodman 1954; Williamson 2013; Russell 2015). We will look at how these methods can be extended to epistemology in general.

Another key issue concerns (f) and what it means to say that we know a belief to be justified or an argument to be valid. Anti-realists might argue that insofar as there is some knowledge involved here it is one of a purely instrumental kind (Kelly 2003), perhaps a kind of know-how that is not fully cognitive but merely dispositional, and hence that is agreeable to the naturalists (Ryle 1946; Boghossian 2003). However, we will argue that once the naturalist edifice in the realms of epistemic normativity starts crumbling, cognitivism about knowledge of facts of justification and validity becomes the frontrunner.

We will defend the view that knowledge of epistemological reality is substantive, as opposed to merely procedural. Since we hold that there are facts about justification and validity, there are epistemological facts to be known. However, our claim that these facts are non-natural raises a question concerning the method by which they are known, since we clearly do not learn epistemological truths by looking, in the way we learn how many socks are in the drawer. Building on Williamson (2007), we will argue that knowledge of normative epistemological truths is a kind of ‘armchair’ knowledge, which involves more than just looking.

Method
This project tackles the big picture, giving a comprehensive defence of robust realism about epistemology. Our methodology therefore involves drawing on many areas of philosophy, including meta-ethics, philosophy of logic, philosophy of mind, philosophy language, metaphysics and epistemology, and applying the issues and insights from these areas to the understanding of epistemological normativity. In particular, we adopt a comparative method, looking at epistemological normativity against the background of systematic investigation of normativity in other domains, such as meta-ethics. The use of this method promises to shed light not only on epistemic normativity, but on normativity in general.

Second, it is worth mentioning that in relation to the semantic aspect of the project in particular, we will use formal methods developed in linguistics and philosophy of language to investigate the semantics of epistemological statements and judgments. The use of formal methods allows for a more rigorous and nuanced debate.

Significance
This project is at the forefront of a new research program in meta-epistemology—one that takes seriously the idea that epistemology is normative, and investigates the foundations of epistemic normativity. This approach to metaepistemology promises to shed new light on the central questions in the semantics, metaphysics, and epistemology of epistemology that we propose to discuss. Moreover, since the idea that epistemology is normative is generally presented as a challenge to robust realism, this project fills a major lacuna in the debate, by systematically defending robust realism with regard to epistemic normativity. In so doing, it makes an original and substantive contribution to the longstanding debate between realism and anti-realism in philosophy.

The significance of the project extends beyond the confines of philosophy. Epistemological relativism and related anti-realist views are highly influential in the humanities in general, so our engagement with these views will be relevant to debates in other disciplines. For instance, the project has a particular significance for science and technology studies, which investigates the epistemology of science, and where relativism has long been the dominant paradigm.

Finally, the project has a potential significance for public discourse in the new age of ‘post-fact’ politics, in which people are swayed more by their emotions and gut instincts than by the facts, evidence or rational arguments. For example, it is directly relevant to popular debates between climate change skeptics and climate scientists, since skeptics and scientists disagree in part over epistemological standards—how much evidence is needed to justify belief in anthropogenic climate change. We argue that it is not a matter of choice, much less politically partisan choice, which standards are appropriate here. We plan to engage with these issues in the popular media both in Sweden and internationally.

Project managers

Members